| name | manas-najimuddeen |
| description | Senior Technical Lead with 13+ years in Python and distributed systems, focused on AI/LLM backends — RAG services, LLM pipelines, and speech-to-text platforms. Use when a project needs backend or AI engineering designed, built, tested, and shipped end to end. Keywords: FastAPI, RAG, pgvector, LLM orchestration, Whisper, ETL, AWS, Cloudflare Workers, Modal. |
Skills
AI / LLM engineering
- LLM integration via OpenAI-compatible SDKs (Claude/Anthropic, Gemini, DeepSeek, xAI Grok)
- Retrieval-augmented generation — chunking, embeddings, pgvector retrieval, relevance validation
- Multi-model pipeline orchestration
- Multimodal (vision and speech) ingestion
- Generative media APIs (fal.ai for image and video, ElevenLabs for speech) assembled with ffmpeg and MoviePy
Backend & APIs
- Python (Django/DRF, FastAPI, Flask)
- REST API and system design
- PostgreSQL/Aurora, pgvector, SQLAlchemy, Alembic
- MongoDB
- Redis (caching, queues, debounced background work)
- Celery and RabbitMQ
- Authentication and authorisation with AWS Cognito — credential and service-to-service flows, token verification, and role/permission modelling
Data & pipelines
- ETL and data pipelines (Airflow, AWS Step Functions)
- Bulk multi-source ingestion
- Speech-to-text (OpenAI Whisper, faster-whisper large-v3)
Cloud & delivery
- AWS (Lambda, Step Functions, Elastic Beanstalk, Fargate, EC2, ALB, RDS/Aurora, CloudWatch, SQS)
- Cloudflare Workers/Pages
- Modal (serverless GPU)
- Docker with multi-stage builds
- NGINX
- Linux/systemd
- CI/CD (GitLab CI, GitHub Actions) with security scanning gates. Develops across Linux, WSL2, macOS and Windows
Operations & reliability
- Sole production ownership
- Deploys, monitoring and alerting, off-box backups, log retention and incident response
- Configuration-as-code with automated drift detection
- Sentry and CloudWatch Logs Insights for error and log observability
AI-assisted development
- Designing and shipping systems by directing agentic coding tools (Claude Code), with a written operating contract for agent sessions and documentation written to be consumed by an agent as well as a human
Client-facing delivery
- On-site solution delivery and integration (Qatar, Saudi Arabia)
- Requirements taken directly from C-level and department heads
- FAT, UAT and handover training
- Multi-vendor multinational programmes
Growth & measurement
- SEO and answer-engine optimisation — content clusters, structured data, internal linking, JS-free rendering for AI crawlers
- Google Search Console and Cloudflare Analytics APIs
- Tracking-pipeline debugging
- Scheduled reporting with week-over-week deltas
Practices
- TDD and PyTest against 90–95% coverage bars
- Clean architecture, SOLID
- Code review
- Agile delivery
Languages
- Python (primary), JavaScript/TypeScript, SQL/NoSQL